SENTENCE PREDICTION IN DEVELOPMENTAL LANGUAGE DISORDER - PROJECT SUMMARY/ABSTRACT Millions of children – 7-12% of the school-age population – have developmental language disorder (DLD), a disorder that affects language learning, comprehension, and use. These children have difficulty with sentence production and comprehension at every stage of their development. These difficulties have major implications for the educational attainment of children with DLD compared to children with typical development (TD). Children with TD and adults make rapid predictions of upcoming words following verbs (e.g., predicting the patient in The dog eats the bone or the agent in The bone is eaten by the dog). Extensive work in children and adults indicates that prediction facilitates sentence comprehension by inducing a state of preparedness and contributes to language development by tuning the language system to the input. Children with DLD have a poorer ability to make sentence predictions, which may compound and result in sentence-comprehension deficits. Prior work on prediction deficits in DLD has focused on broad sentence characteristics like typicality or broad participant characteristics like vocabulary test scores. However, studies with typical individuals have identified more specific sentence- and participant-level contributors to sentence prediction. These factors have not been systematically explored in children with DLD. We explore three separate hypotheses concerning factors that affect prediction in DLD. First, sentence- and participant-level properties affect prediction in children with DLD. Second, children with DLD lack robust representations of the underlying linguistic knowledge needed to predict, particularly abstract semantic features. Third, children with DLD have differences in event processing that relate to sentence prediction skill. We investigate these hypotheses in 5-7-year-old school-age children with DLD across three Aims. Aim 1: Measure the effect of sentence- and participant-level properties on sentence prediction. We will measure two participant-level cognitive factors (processing speed, verbal working memory) and their effect on prediction of agents and patients in sentences varying across two properties (syntactic complexity, semantic competition). Aim 2: Measure linguistic knowledge that underlies sentence prediction. We will measure knowledge of agent-verb and verb-patient cooccurrences (e.g., dog-bite, eat-apple) and knowledge of verb-specific semantic features (e.g., throw-<round object>). Aim 3: Measure event- processing skills that underlie sentence prediction. We will measure how children with DLD categorize and attend to agents/patients in visual scenes. Impact: This project has the promise to be highly impactful. First, it bridges disparate literatures on language processing in adults, children with TD, and children with DLD, providing clarity about predictive processing in DLD. Second, it may influence intervention approaches by identifying areas of strength and need in children with DLD. Third, it sets the stage for larger-scale longitudinal work examining effects of early prediction ability, language knowledge, and event processing on later sentence comprehension.
